A comprehensive sampling network was implemented in the Alamosa River Basin from 1995 to 1997 to address data gaps identified as part of the ecological risk assessment of the Summitville Superfund site. Aluminum, copper, iron, and zinc were identified as the constituents of concern for the risk assessment. Water-quality samples were collected at six sites on the Alamosa River and Wightman Fork by automatic samplers. Several discrete (instantaneous) samples were collected over 24 hours at each site during periods of high diurnal variations in streamflow (May through September). The discrete samples were analyzed individually and duplicate samples were composited to produce a single sample that represented the daily-mean concentration. The diurnal variations in concentration with respect to the theoretical daily-mean concentration (maximum minus minimum divided by daily mean) are presented. Diurnal metal concentrations were highly variable in the Alamosa River and Wightman Fork. The concentration of a metal at a single site could change by several hundred percent during one diurnal cycle. The largest percent change in metal concentrations was observed for aluminum and iron. Zinc concentrations varied the least of the four metals. No discernible or predictable pattern was indicated in the timing of the daily mean, maximum, or minimum concentrations. The percentage of discrete sample concentrations that varied from the daily-mean concentration by thresholds of plus or minus 10, 25, and 50 percent was evaluated. Between 50 and 75 percent of discrete-sample concentrations varied from the daily-mean concentration by more than plus or minus 10 percent. The percentage of samples exceeding given thresholds generally was smaller during the summer period than the snowmelt period. Sampling strategies are critical to accurately define variability in constituent concentration, and conversely, understanding constituent variability is important in determining appropriate sampling strategies. During nonsteady-state periods, considerable errors in estimates of daily-mean concentration are possible if based on one discrete sample. Flow-weighting multiple discrete samples collected over a diurnal cycle provides a better estimate of daily-mean concentrations during nonsteady-state periods.